He turned his laptop round and showed me a chart. Three coloured lines, one of them heading downwards, and a question: should he be worried?
He had been training seriously for six years and looking at that chart most days for two of them. When I asked him what the blue line actually measured, he said fitness, which is what the software calls it. When I asked him fitness of what, measured how, he did not know.
That is not his fault. These numbers get presented as though they are self-explanatory, and then people build their training year around them without ever being told what is underneath.
It is simpler than it looks. There is one number per session, and three running averages of that number. That is the whole system.
The chart is a good description of what your legs have done. It is not a description of you.
TSS: One Number for One Session
Training Stress Score takes how long you went and how hard you went relative to your own threshold, and turns it into a single figure.
The anchor is deliberately simple: one hour at your threshold is 100. Two hours at a genuinely easy pace might be 90. Forty minutes of hard intervals might be 70. The scale is arbitrary in the way that a metre is arbitrary, and useful for exactly the same reason. It lets you compare a swim, a turbo session and a long run in one currency.
Two things follow. First, TSS is only as good as the threshold you set it against. If your FTP is three months stale and you have got fitter, every session is scoring lower than it should, and your chart will tell you that you are losing fitness while you are gaining it. Test properly, and update the number.
Second, an hour is an hour. TSS cannot tell the difference between an hour at threshold on a Tuesday when you slept nine hours, and the same hour on a Tuesday when you did not sleep at all. Hold on to that, because it is the entire limitation of the system and we will come back to it.
CTL, ATL and TSB
CTL is a rolling average of your daily TSS over about six weeks. It moves slowly, because it is meant to. It represents the work your body has actually absorbed and adapted to, which is why the software labels it fitness. One enormous weekend barely shifts it. Four consistent months move it a long way.
ATL is the same calculation over about a week. It moves fast. It spikes after a big block and drops away within days of going easy. That is fatigue.
TSB is simply CTL minus ATL. Fitness minus fatigue. The software calls it form. When it is negative you are carrying more fatigue than your current fitness comfortably absorbs, which is precisely what a build block should look like. When it is positive you are fresher than your baseline, which is what a taper is engineering.
That is it. Everything else is presentation.
The Number Is Meaningless. The Shape Is Not.
The most common mistake is comparing your CTL with somebody else's. It is calculated from your threshold, so it is a measure of your training relative to your own capacity. A CTL of 70 in one athlete and 70 in another describe two completely different training loads in absolute terms.
What is worth reading is the shape of your own line over months. Is it climbing, flat, or falling? Was it higher this time last year at the same point in the season? Did it fall off a cliff in a week you now remember as the week everything went wrong at work?
Used that way it is genuinely valuable, because it is a memory. Human beings are extremely bad at remembering how much they trained four months ago, and consistently generous to themselves about it.
How Fast You Can Build
This is the part with real practical consequence.
Building CTL too quickly is one of the more reliable ways to get injured. A reasonable working range on the bike is around three to eight points a week. For running it should be more conservative, nearer three to five, and the reason is not cardiovascular. Your heart and lungs will happily accept far more than that. Your tendons and bones will not, and they adapt on a slower clock than your engine does.
The related idea is the ratio between your acute and chronic load. Keep the week you are in somewhere between about 0.8 and 1.3 times your recent baseline and you are in sensible territory. Push it past about 1.5 and injury risk climbs steeply. That single ratio explains most of the training errors I see, and almost all of them happen in the same circumstances: an athlete gets a free week, feels good, and does in seven days what they normally do in twelve.
For race day, you generally want to arrive somewhere between about minus ten and plus twenty five on TSB. A taper works by dropping ATL quickly while CTL, which moves slowly, stays where it is. That gap is the freshness. Taper too long and CTL starts falling too, which is how people arrive at a start line rested and flat.
What the Chart Cannot See
Now the important part, and the reason I am careful about handing these tools to athletes without a conversation.
TSS counts what you did with your legs. It does not count the meeting that went badly, the four hours of broken sleep because a child was ill, the funeral, the deadline, the move, the argument in the car park. None of that appears anywhere on the chart, and all of it lands in the same physiological account as your training.
So you can be sitting at a perfectly reasonable TSB of minus five, entirely on plan, and be genuinely unfit to train, and the software will show you three tidy lines suggesting everything is fine. I have watched athletes override a body that was shouting at them because a number said they were fresh.
Garmin's version of all this uses different names, training status, acute load, load ratio, and the same underlying maths. It has the additional quirk that its baselines shift as it learns you, so the labels can change without your training changing at all.
Whatever the branding, the rule is the same one. The chart is a record of load, not a verdict on readiness. Put it next to your resting heart rate, your heart rate variability trend and the first ten minutes of your warm up, and when the chart and your body disagree, back the body. The chart has never once been right about that.
Three Questions Is All You Need
Open it once a week, not once an hour, and ask three things.
Am I building, holding or recovering, and is that what this block was supposed to be doing? Is the ramp inside a sane range, or did I add a third of my weekly load because the weather was good? And does the picture match how I actually feel?
If you cannot answer the third one, you are not training with data. You are being trained by it.
When You Miss a Week
Something will go wrong. A chest infection, a work trip, a child who brings something home from school. The chart will show your CTL falling and it will feel like watching money leave an account.
Here is what is actually happening. CTL is a six week average, so one missed week moves it far less than the drop on the screen suggests, and what you lose in seven days you get back in roughly the same seven days. Two weeks off will cost you a handful of points. The fitness underneath, the mitochondria and the capillaries and the years of aerobic work, is not stored in that line.
What does real damage is the response. An athlete watches the line fall for ten days and then tries to buy it back in one weekend, which sends the acute ratio through 1.5 and produces the injury that costs them two months rather than two weeks.
Come back at the load you left at, not the load you feel you owe. The line will follow. It always does.
The related trap is treating CTL as a score to be maximised. It is a description, not a target. An athlete who chases a bigger number for its own sake will get there, usually by adding volume they cannot absorb, and will arrive at their race with an impressive chart and nothing in their legs. The number exists to describe training that has a purpose. It is not the purpose.
The Challenge
Pull up your last twelve weeks and find the steepest week on the chart.
Now look at what happened in the fortnight after it. The illness, the niggle, the flat sessions you blamed on motivation.
It is usually there in the shape, and you usually already knew.